PCAST Co-Chair Criticizes OpenAI and Anthropic Over AI Slowdown Calls
David Sacks, Co-Chair of the US President's Council of Advisors on Science and Technology, publicly criticized OpenAI and Anthropic for using calls to slow down frontier AI development as a tactic for regulatory capture. He argued that if these leading labs feel their models are dangerous, they should voluntarily pause development rather than demanding antitrust exemptions or government-backed cartels. The commentary highlights a growing divide in tech policy over whether AI safety concerns are genuine or being weaponized by industry leaders to erect regulatory barriers against smaller competitors. It signals potential government resistance against granting liability waivers or custom antitrust protections to established AI duopolies. Sacks specifically questioned the independence of safety evaluation bodies like METR, pointing out potential conflicts of interest due to ties with Anthropic investors and employees. He also noted that prioritizing reliability and predictability over raw capabilities is standard business risk management to avoid catastrophic cyberattack liabilities, rather than pure altruism.
## BACKGROUND
Frontier AI safety evaluation often relies on independent organizations like METR (Model Evaluation and Threat Research), a non-profit institute assessing agentic AI capabilities and catastrophic risks. Meanwhile, concerns over frontier models are heavily tied to concepts like recursive self-improvement, where an AI autonomously rewrites its own code to enhance its capabilities, potentially leading to uncontrollable acceleration in AI power.